Researchers at Prof. Itamar Lensky’s Remote Sensing and Geographic Information Systems Laboratory are advancing fire-risk mapping by incorporating satellite-based metrics that evaluate long-term vegetation status and dryness. Their study, utilizing machine learning algorithms like Logistic Regression, Random Forest, and XGBoost, specifically analyzed the 2007 wildfires in Greece.

The findings, recently detailed in their research, show that the XGBoost algorithm, which handles variable interactions and non-linear effects, most effectively improved fire-risk map quality. This method, particularly with indices such as woody vegetation density and long-term drought effects, offers a more accurate approach to mapping fire-prone areas than traditional methods. Such enhanced mapping can significantly aid in both fire prevention and firefighting by providing a deeper understanding of plant dynamics and potential fire behavior.

Click here to learn more.